CN110718301A - Alzheimer disease auxiliary diagnosis device and method based on dynamic brain function network - Google Patents
Alzheimer disease auxiliary diagnosis device and method based on dynamic brain function network Download PDFInfo
- Publication number
- CN110718301A CN110718301A CN201910916563.8A CN201910916563A CN110718301A CN 110718301 A CN110718301 A CN 110718301A CN 201910916563 A CN201910916563 A CN 201910916563A CN 110718301 A CN110718301 A CN 110718301A
- Authority
- CN
- China
- Prior art keywords
- nuclear magnetic
- magnetic resonance
- dynamic
- functional nuclear
- feature
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 230000003925 brain function Effects 0.000 title claims abstract description 131
- 238000003745 diagnosis Methods 0.000 title claims abstract description 62
- 208000024827 Alzheimer disease Diseases 0.000 title claims abstract description 40
- 238000000034 method Methods 0.000 title claims abstract description 36
- 210000004556 brain Anatomy 0.000 claims abstract description 107
- 239000011159 matrix material Substances 0.000 claims abstract description 38
- 238000012549 training Methods 0.000 claims abstract description 27
- 238000005259 measurement Methods 0.000 claims abstract description 26
- 238000012216 screening Methods 0.000 claims abstract description 19
- 238000007781 pre-processing Methods 0.000 claims abstract description 17
- 238000002599 functional magnetic resonance imaging Methods 0.000 claims abstract description 5
- 238000005481 NMR spectroscopy Methods 0.000 claims description 195
- 238000012937 correction Methods 0.000 claims description 21
- 238000001914 filtration Methods 0.000 claims description 18
- 238000004422 calculation algorithm Methods 0.000 claims description 13
- 230000004886 head movement Effects 0.000 claims description 13
- 239000003638 chemical reducing agent Substances 0.000 claims description 11
- 238000012706 support-vector machine Methods 0.000 claims description 9
- 230000009467 reduction Effects 0.000 claims description 8
- 238000000605 extraction Methods 0.000 claims description 6
- 238000010276 construction Methods 0.000 claims description 5
- 230000011218 segmentation Effects 0.000 claims description 5
- 230000006870 function Effects 0.000 claims description 4
- 238000009499 grossing Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 2
- 239000000284 extract Substances 0.000 claims 1
- 230000007547 defect Effects 0.000 abstract description 3
- 230000000694 effects Effects 0.000 abstract description 3
- 230000003068 static effect Effects 0.000 abstract description 3
- 230000007177 brain activity Effects 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 5
- 210000002569 neuron Anatomy 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000007635 classification algorithm Methods 0.000 description 2
- 238000004195 computer-aided diagnosis Methods 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 230000002123 temporal effect Effects 0.000 description 2
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000008280 blood Substances 0.000 description 1
- 210000004369 blood Anatomy 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000000739 chaotic effect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000005714 functional activity Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000003012 network analysis Methods 0.000 description 1
- 238000002610 neuroimaging Methods 0.000 description 1
- 229910052760 oxygen Inorganic materials 0.000 description 1
- 239000001301 oxygen Substances 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000000284 resting effect Effects 0.000 description 1
- 238000012731 temporal analysis Methods 0.000 description 1
- 238000000700 time series analysis Methods 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
- G06T7/33—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
- G06T7/38—Registration of image sequences
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10088—Magnetic resonance imaging [MRI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30016—Brain
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Biomedical Technology (AREA)
- Medical Informatics (AREA)
- Public Health (AREA)
- Primary Health Care (AREA)
- General Health & Medical Sciences (AREA)
- Epidemiology (AREA)
- Pathology (AREA)
- Databases & Information Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Magnetic Resonance Imaging Apparatus (AREA)
Abstract
The invention discloses an Alzheimer disease auxiliary diagnosis device and method based on a dynamic brain function network. The diagnosis device comprises an fMRI data preprocessing unit, a dynamic brain function network building unit, a feature generating unit for training and an svm classification auxiliary diagnosis unit, and the use method of the diagnosis device comprises the following steps: firstly, preprocessing images, then constructing a dynamic brain network, secondly, calculating node measurement of the segmented brain network, forming a time sequence by each node measurement through a time sequence generator, then, extracting characteristics for the formed time sequence through a characteristic extractor, splicing the filtered characteristics into a matrix through a characteristic filter, screening through a characteristic screening device, finally, carrying out classification training on data through a data training device, and finally, realizing the diagnosis on the Alzheimer's disease through an auxiliary diagnosis device. The method overcomes the defect that the static brain function network can not express dynamic information, and has better effect of serving medical auxiliary diagnosis.
Description
Technical Field
The invention belongs to the technical field of computer-aided diagnosis, and relates to an Alzheimer disease auxiliary diagnosis device and method based on a support vector machine classification algorithm, in particular to an Alzheimer disease auxiliary diagnosis device and method based on a dynamic brain function network.
Background
In recent years, the rapid progress of neuroimaging technology, especially functional imaging, provides a corresponding technology for researching the functions of each brain region of a patient suffering from Alzheimer's disease. The functional magnetic resonance imaging is taken as one of mature functional imaging detection technologies, and has the advantages of non-invasive monitoring of brain functions and activities and high spatial and temporal resolution. Measuring the correlation between various regions of the brain using a blood oxygen level dependent based method has proven to be a powerful tool for studying functional tissues of the brain. In resting fMRI, it can be seen that there is ordered functional activity in the various brain regions of the human brain at this time.
The brain is a dynamic structure, the connection between neurons in the brain changes along with the change of time, and the transient characteristics of each time period in the fMRI data can be better analyzed by analyzing the brain function connection network based on a dynamic network method. Through the construction and analysis of dynamic brain function networks, the activity state of the brain and the interaction between individual neurons or brain regions can be better described.
Disclosure of Invention
Aiming at the defects of the prior art, the invention aims to provide an auxiliary diagnostic device and method for Alzheimer's disease based on a dynamic brain function network, which classify a functional Magnetic Resonance image (fMRI) by using a support vector machine classification algorithm technology and a dynamic brain function network construction technology so as to effectively diagnose the Alzheimer's disease, can acquire more dynamic brain activity information and can accurately judge the difference between brain networks in different states, and the specific scheme is as follows:
an auxiliary diagnostic device for Alzheimer disease based on dynamic brain function networks comprises an fMRI data preprocessing unit, a dynamic brain function network constructing unit, a feature unit for generating training and an svm classification auxiliary diagnostic unit, wherein I acquired functional nuclear magnetic resonance images to be tested are subjected to time dimension division through the fMRI data preprocessing unit to obtain i preprocessed standard functional nuclear magnetic resonance images, the i preprocessed standard functional nuclear magnetic resonance images are subjected to time dimension division through the dynamic brain function network constructing unit to construct i s dynamic brain function networks of the i functional nuclear magnetic resonance images, the constructed i s dynamic brain function networks are subjected to transverse feature extraction of each dynamic brain function network through the feature unit for generating training, and the dynamic features extracted from the i s dynamic brain function networks and filtered by a feature filter are spliced into one dynamic special feature And (4) characterizing the matrix, and finally, screening characteristics by using a Fisher algorithm through the svm classification auxiliary diagnosis unit, screening representative characteristics, and training for auxiliary diagnosis.
The fMRI data preprocessing unit comprises a time slice corrector, a head movement corrector, a space standardizer and a smooth noise reducer, firstly, the acquired i functional nuclear magnetic resonance images to be detected are subjected to time slice correction through the time slice corrector to obtain i corrected functional nuclear magnetic resonance images, then the i time-corrected functional nuclear magnetic resonance images are corrected in a head movement manner through the head movement corrector to obtain i head movement-corrected functional nuclear magnetic resonance images, secondly, spatially standardizing the i functional nuclear magnetic resonance images subjected to the head motion correction through the spatial standardizer to obtain i functional nuclear magnetic resonance images subjected to spatial standardization, and finally, smoothly denoising the i functional nuclear magnetic resonance images subjected to the spatial standardization through the smooth denoiser to obtain i standard functional nuclear magnetic resonance images;
the time slice corrector is used for performing time slice correction on the input I functional nuclear magnetic images to be detected to obtain I corrected functional nuclear magnetic resonance images (I) of time slices-1,I-2,I-3,…,I-i) Wherein i represents the number of the selected functional nuclear magnetic resonance images to be detected;
the head movement corrector is used for correcting the functional nuclear magnetic image (I) after the I time slices-1,I-2,I-3,…,I-i) Performing head motion correction to obtain i functional nuclear magnetic resonance images (H) after head motion correction-1,H-2,H-3,...,H-i)；
The spatial normalizer is used for correcting i functional nuclear magnetic images (H) after head movement-1,H-2,H-3,...,H-i) Spatially normalizing to obtain i spatially normalized functional nuclear magnetic resonance images (F)-1,F-2,F-3,...,F-i)；
The smoothing noise reducer is used for normalizing i spatially functional nuclear magnetic resonance images (F)-1,F-2,F-3,…,F-i) Performing smooth noise reduction to obtain i standard functional nuclear magnetic resonance images (S)-1,S-2,S-3,…,S-i)。
The method comprises the steps that a dynamic brain function network building unit comprises a template matcher, a time sequence divider and a brain network builder, wherein each standard functional nuclear magnetic resonance image in i standard functional nuclear magnetic resonance images is matched with a standard matching template with the specification of m in an interested area through the template matcher, so that each matched functional nuclear magnetic resonance image comprises m brain areas, the time sequence of each brain area is divided through the time sequence divider of the matched functional nuclear magnetic resonance image to obtain s sections of instant time sequences, and finally the i x s dynamic brain function networks of the i divided functional nuclear magnetic resonance images are built through the brain network builder of the s sections of instant time sequences obtained through division;
the template matcher is used for matching each standard functional nuclear magnetic resonance image with a standard matching template with the specification of m, each matched functional nuclear magnetic resonance image comprises m brain areas, and the i matched functional nuclear magnetic resonance images are expressed as (A)-1,A-2,A-3,...,A-i)；
The time sequence divider is used for dividing the time sequence of m brain regions in each matched functional nuclear magnetic resonance image into s sections of instant time sequences, and each section of instant time sequence represents instant information (T) of one brain region-1,T-2,...,T-s) I segmented functional NMR images are expressed as
Wherein the value range of s is determined according to a preset segmentation interval;
the brain network builder is used for building s dynamic brain function networks from s sections of instantaneous time sequences in each segmented functional nuclear magnetic resonance image, and the i segmented functional nuclear magnetic resonance images obtain i x s dynamic brain function networks
The feature unit for training is generated and comprises a node measurement generator, a time sequence generator, a feature extractor and a feature filter, wherein i x s dynamic brain function networks for constructing i segmented functional nuclear magnetic resonance images output by the dynamic brain function network unit are used for calculating the node measurement of the i x s dynamic brain function networks through the node measurement generator, then each node measurement in the i x s dynamic brain function networks is formed into a time sequence through the time sequence generator, a new feature value is extracted for the time sequence formed by each node measurement through the feature extractor, and finally all the extracted new feature values are filtered through the feature filter and then spliced into a dynamic feature matrix;
the node metric generator is used for generating node metrics of i-s dynamic brain function networks and calculating the characteristic value of z dynamic characteristics to be represented as
The z dynamic features comprise i s x global features and i s m y local features, namely z i s x i m y, x represents the number of global features calculated by each dynamic brain function network, and y represents the number of local features calculated by each dynamic brain function network;
the time sequence generator is used for generating a time sequence from each group of obtained characteristic values, and z/s time sequences formed by i-s dynamic brain function networks are represented as
Each group of feature values comprises global feature values of each group and local feature values of each group, each global feature value of each group comprises feature values of s dynamic brain function networks of each global feature in each segmented functional nuclear magnetic resonance image, and each local feature value of each group comprises feature values of s dynamic brain function networks obtained by grouping each local feature of s m y local features in each segmented functional nuclear magnetic resonance image according to m brain regions;
the feature extractor is used for extracting features again for the time series generated by each group of features based on the time series entropy of the wavelet to obtain new feature values of z/s features
The feature filter is used for filtering i s m y local features in the node metric, firstly, the degree of each brain area in each matched functional nuclear magnetic resonance image is calculated by a degree method in the node metric, and the average value of m brain area degrees in each matched functional nuclear magnetic resonance image is calculated
Then calculating the standard deviation sigma of m brain areas in each matched functional nuclear magnetic resonance image, and then enabling the degree in the m brain areas to belong to an interval
Filtering the brain areas to obtain key brain areas, finally splicing new eigenvalues generated by N x local characteristics of the N brain areas and i x global characteristics obtained by filtering the matched functional nuclear magnetic resonance images into a dynamic characteristic matrix through a characteristic filter,
wherein n isjThe number of key brain areas in the jth matched functional nuclear magnetic resonance image is shown, and the number of matched functional nuclear magnetic resonance images is shown as i.
The svm classification auxiliary diagnosis unit comprises a feature filter, a data trainer and an auxiliary diagnostor, and the dynamic feature matrix output by the feature unit for training is firstly subjected to dynamic feature screening by the feature filter and a Fisher algorithm, then the classification training is carried out by the data trainer, and finally the diagnosis of the Alzheimer disease is realized by the auxiliary diagnostor;
the feature filter is used for scoring N x y + i x features in the dynamic feature matrix by using a Fisher algorithm, sorting the N x y + i x features according to the sequence of scores from top to bottom, screening the top w features with high scores as the most representative features, and determining w according to actual conditions;
the data trainer is used for training the screened first w most representative features according to the functional nuclear magnetic resonance image to obtain a classifier in the classification of the support vector machine;
the auxiliary diagnosis device is used for carrying out auxiliary diagnosis on the Alzheimer disease according to the trained classifier.
A method for using an Alzheimer disease auxiliary diagnosis device based on a dynamic brain function network comprises the following steps:
step 1: preprocessing a functional nuclear magnetic resonance image;
step 2: constructing a dynamic brain function network by utilizing the preprocessed image;
and step 3: calculating node measurement of the dynamic brain function networks, extracting dynamic characteristics of each dynamic brain function network, and splicing characteristic values of the filtered dynamic characteristics into a dynamic characteristic matrix;
and 4, step 4: and performing auxiliary diagnosis on the Alzheimer disease by using the generated dynamic feature matrix.
The step 1 of preprocessing the functional nuclear magnetic resonance image comprises the following steps:
1.1) the acquired I functional nuclear magnetic resonance images to be detected are subjected to time slice correction by the time slice corrector to obtain I time slice corrected functional nuclear magnetic resonance images (I)-1,I-2,I-3,…,I-i) Wherein i represents the number of the selected functional nuclear magnetic resonance images to be detected;
1.2) performing head motion correction on the i time-corrected functional nuclear magnetic resonance images through the head motion corrector to obtain i head motion-corrected functional nuclear magnetic resonance images (H)-1,H-2,H-3,...,H-i)；
1.3) the i functional nuclear magnetic resonance images after the head movement correction are subjected to spatial standardization through the spatial standardizer to obtain i functional nuclear magnetic resonance images (F) after the spatial standardization-1,F-2,F-3,...,F-i)；
1.4) the i spatially normalized functional nuclear magnetic resonance images are subjected to smooth noise reduction by the smooth noise reducer to obtain i standard functional nuclear magnetic resonance images (S)-1,S-2,S-3,...,S-i)。
The step 2 of constructing a dynamic brain function network by utilizing the preprocessed image comprises the following steps:
2.1) combining i standard functional NMRMatching each standard functional nuclear magnetic resonance image in the images with a standard matching template with the specification of m through the template matcher to perform region-of-interest matching, wherein each matched functional nuclear magnetic resonance image comprises m brain areas, and obtaining i matched functional nuclear magnetic resonance images (A)-1,A-2,A-3,...,A-i)；
2.2) the i matched functional nuclear magnetic resonance images are used for segmenting the time sequence of each brain area in each matched functional nuclear magnetic resonance image through the time sequence segmenter to obtain s sections of instant time sequences, and each section of instant time sequence represents instant information (T) of one brain area-1,T-2,...,T-s) I segmented functional NMR images are expressed as
Wherein the value range of s is determined according to a preset segmentation interval;
2.3) constructing i x s dynamic brain function networks of i segmented functional nuclear magnetic resonance images by the segmented s-segment instantaneous time sequence through the brain network constructor
Step 3, calculating node measurement for the dynamic brain function networks, extracting the dynamic characteristics of each dynamic brain function network, and then splicing the characteristic values of the filtered dynamic characteristics into a dynamic characteristic matrix, wherein the method comprises the following steps:
3.1) generating node measurement of i x s dynamic brain function networks of the i segmented functional nuclear magnetic resonance images through the node measurement generator, and calculating to obtain the characteristic value of z dynamic characteristics to be represented as
The z dynamic features include i s x global features and i s m y local features, i.e., z i x + i x m y, x represents the number of global features calculated by each dynamic brain function network, and y represents the work of each dynamic brainThe number of local features that can be computed over a network;
3.2) generating a time sequence by the time sequence generator for each group of characteristic values, and then expressing z/s time sequences formed by i × s dynamic brain function networks as
Each group of feature values comprises global feature values of each group and local feature values of each group, each global feature value of each group comprises feature values of s dynamic brain function networks of each global feature in each segmented functional nuclear magnetic resonance image, and each local feature value of each group comprises feature values of s dynamic brain function networks obtained by grouping each local feature of s m y local features in each segmented functional nuclear magnetic resonance image according to m brain regions;
3.3) the time series generated by each group of features is subjected to feature extraction again through the wavelet-based time series entropy in the feature extractor to obtain new feature values of z/s features
3.4) filtering i s m y local features in the node measure by a feature filter, firstly calculating the degree of each brain area in each matched functional nuclear magnetic resonance image by a degree method in the node measure, and calculating the average value of the degrees of the m brain areas in each matched functional nuclear magnetic resonance image
Then calculating the standard deviation sigma of m brain areas in each matched functional nuclear magnetic resonance image, and then enabling the degree in the m brain areas to belong to an interval
Filtering the brain areas into key brain areas, and finally splicing new eigenvalues generated by N x y local characteristics of the N brain areas and i x global characteristics obtained by filtering the matched functional nuclear magnetic resonance images through a characteristic filterFollowed by a matrix of dynamic characteristics,
wherein n isjThe number of key brain areas in the jth matched functional nuclear magnetic resonance image is represented, and i represents the number of matched functional nuclear magnetic resonance images;
the step 4 of performing auxiliary diagnosis on the Alzheimer disease by using the generated dynamic feature matrix comprises the following steps:
4.1) scoring N x y + i x features in the dynamic feature matrix extracted by the feature filter through a Fisher algorithm in a feature filter, sorting the features according to the sequence of scores from top to bottom, screening the top w features with high scores as the most representative features, and determining w according to the actual condition;
4.2) training the first w most representative features screened out by a support vector machine in a data trainer to obtain a classifier;
4.3) using an auxiliary diagnosis device to perform auxiliary diagnosis of the Alzheimer disease through the obtained classifier.
The invention has the beneficial effects that:
the invention relates to an Alzheimer disease auxiliary diagnosis device and method based on a dynamic brain function network, which overcome the defect that the conventional static brain function network cannot express dynamic information, enable dynamic analysis of brain activity information to be possible, fully exert signal information of a functional nuclear magnetic resonance image and play a better role in medical auxiliary diagnosis service.
Drawings
Fig. 1 is a block diagram of an alzheimer's disease auxiliary diagnosis apparatus based on a dynamic brain function network in an embodiment of the present invention.
Fig. 2 is a flowchart of a dynamic brain function network construction method in an embodiment of the present invention.
Fig. 3 is a flowchart of a method for dynamic information feature extraction and filtering and diagnosis assistance in an embodiment of the present invention.
Detailed Description
The following is a detailed description of the technical solution of the present invention with reference to the accompanying drawings.
In fMRI data, signals at different times include brain activity information, but existing research is based on static network analysis of a complex network, neglecting that the brain is a dynamic structure, and connections between neurons in the brain change with time, so that a dynamic brain function network needs to be constructed to more comprehensively describe the brain activity information, and calculate characteristics of node measurement characteristic values of each dynamic brain function network. Therefore, a design method based on a dynamic brain function network is proposed, which considers instantaneous differences of the brain network in a plurality of time periods and the structure of the brain network, and is used for the auxiliary diagnosis of the alzheimer disease based on the support vector machine.
As shown in fig. 1, the dynamic brain function network-based alzheimer disease auxiliary diagnosis device includes an fMRI data preprocessing unit, a dynamic brain function network constructing unit, a feature generating unit for training, and an svm classification auxiliary diagnosis unit, wherein 606 preprocessed standard functional nuclear magnetic resonance images are obtained from 606 acquired to-be-detected functional nuclear magnetic resonance images through the fMRI data preprocessing unit, the preprocessed 606 standard functional nuclear magnetic resonance images are divided in a time dimension through the dynamic brain function network constructing unit, 15150 dynamic brain function networks of the 606 functional nuclear magnetic resonance images are constructed, the constructed 15150 dynamic brain function networks are transversely extracted from the feature generating unit for training through the dynamic brain function networks, and finally, the dynamic features extracted from the 15150 dynamic brain function networks and filtered by a feature filter are spliced into a dynamic feature matrix And finally, performing feature screening by using a Fisher algorithm through the svm classification auxiliary diagnosis unit, screening representative features, and training for auxiliary diagnosis.
The fMRI data preprocessing unit comprises a time slice corrector, a head movement corrector, a space standardizer and a smooth noise reducer, firstly, 606 acquired functional nuclear magnetic resonance images to be detected are subjected to time slice correction through the time slice corrector to obtain 606 corrected functional nuclear magnetic resonance images, then, performing head motion correction on the 606 time-corrected functional nuclear magnetic resonance images through the head motion corrector to obtain 606 head motion-corrected functional nuclear magnetic resonance images, secondly, the 606 functionally-magnetic-resonance images subjected to the dynamic correction are subjected to spatial standardization through the spatial standardizer to obtain 606 functionally-magnetic-resonance images subjected to spatial standardization, and finally the 606 functionally-magnetic-resonance images subjected to the spatial standardization are subjected to smooth noise reduction through the smooth noise reducer to obtain 606 standard functionally-magnetic-resonance images;
the time slice corrector is used for performing time slice correction on the input 606 functional nuclear magnetic images to be detected to obtain 606 time slice corrected functional nuclear magnetic resonance images (I)-1,I-2,I-3,...,I-606)；
The head movement corrector is used for correcting the 606 time slices of the functional nuclear magnetic images (I)-1,I-2,I-3,...,I-606) Performing head movement correction to obtain 606 head movement corrected functional nuclear magnetic resonance images (H)-1,H-2,H-3,...,H-606)；
The space normalizer is used for correcting 606 head movements of the functional nuclear magnetic images (H)-1,H-2,H-3,...,H-606) Spatial normalization is carried out to obtain 606 spatially normalized functional nuclear magnetic resonance images (F)-1,F-2,F-3,...,F-606)；
The smoothing noise reducer is used for normalizing 606 spatially normalized functional nuclear magnetic resonance images (F)-1,F-2,F-3,...,F-606) Carrying out smooth noise reduction to obtain 606 standard functional nuclear magnetic resonance images (S)-1,S-2,S-3,...,S-606)。
Firstly, matching each standard functional nuclear magnetic resonance image in 606 standard functional nuclear magnetic resonance images with a standard power-264 template of 264 brain regions through the template matcher to obtain an interested region, wherein each matched functional nuclear magnetic resonance image comprises 264 brain regions, then segmenting the time sequence of each brain region through the time sequence divider to obtain 25 sections of instant time sequences, and finally constructing 15150 dynamic brain functional networks of the 606 segmented functional nuclear magnetic resonance images through the brain network builder through the 25 sections of instant time sequences obtained by segmentation;
the template matcher is used for matching each standard functional nuclear magnetic resonance image with a standard power-264 template of 264 brain regions, each matched functional nuclear magnetic resonance image comprises 264 brain regions, and 606 matched functional nuclear magnetic resonance images are represented as (A)-1,A-2,A-3,...,A-606)；
The time sequence divider is used for dividing the time sequence of 264 brain regions in each matched functional nuclear magnetic resonance image into 25 instantaneous time sequences, and each instantaneous time sequence represents the instantaneous information (T) of one brain region-1,T-2,...,T-25) And 606 segmented functional nuclear magnetic resonance images are represented as
The brain network builder is used for building 25 dynamic brain function networks from 25 segments of instantaneous time sequences in each segmented functional nuclear magnetic resonance image, and then 15150 dynamic brain function networks are obtained from 606 segmented functional nuclear magnetic resonance images
The feature unit for training generation comprises a node measurement generator, a time sequence generator, a feature extractor and a feature filter, wherein 15150 dynamic brain function networks for constructing 606 segmented functional nuclear magnetic resonance images output by the dynamic brain function network unit are used for calculating 15150 node measurements of the dynamic brain function networks through the node measurement generator, then each node measurement in the 15150 dynamic brain function networks is formed into a time sequence through the time sequence generator, a new feature value is extracted for the time sequence formed by each node measurement through the feature extractor, and finally all the extracted new feature values are filtered through the feature filter and spliced into a dynamic feature matrix;
the node metric generator is used for generating 15150 node metrics of the dynamic brain function network and calculating 28057800 characteristic values of dynamic characteristics to be represented as
The 28057800 dynamic features comprise 60600 global features and 27997200 local features, the number of the global features calculated by each dynamic brain function network is 4, and the number of the local features calculated by each dynamic brain function network is 7;
the time sequence generator is used for generating a time sequence from each group of obtained characteristic values, and 1122312 time sequences formed by 15150 dynamic brain function networks are represented as
Each group of feature values comprises global feature values of each group and local feature values of each group, each global feature value of each group comprises 25 dynamic brain function network feature values of each global feature in each segmented functional nuclear magnetic resonance image, and each local feature value of each group comprises 25 dynamic brain function network feature values obtained by grouping each local feature of 46200 local features in each segmented functional nuclear magnetic resonance image according to 264 brain regions;
the feature extractor is used for extracting features again for the time series generated by each group of features based on the time series entropy of the wavelet to obtain new feature values of 1122312 features
The feature filter is configured to filter 27997200 local features in the node metric, and first calculate a degree of each brain region in each matched functional nuclear magnetic resonance image by using a degree method in the node metric, and calculate an average value of the degrees of the 264 brain regions in each matched functional nuclear magnetic resonance image
Then calculating the standard deviation sigma of 264 brain areas in each matched functional nuclear magnetic resonance image, and then enabling the degree in the 264 brain areas to belong to an interval
And finally, splicing 59388 new characteristic values generated by local characteristics of 8484 brain areas obtained by filtering 606 matched functional nuclear magnetic resonance images and 2424 new characteristic values generated by global characteristics into a dynamic characteristic matrix through a characteristic filter.
The svm classification auxiliary diagnosis unit comprises a feature filter, a data trainer and an auxiliary diagnostor, and the dynamic feature matrix output by the feature unit for training is firstly subjected to dynamic feature screening by the feature filter and a Fisher algorithm, then the classification training is carried out by the data trainer, and finally the diagnosis of the Alzheimer disease is realized by the auxiliary diagnostor;
the feature filter is used for scoring 61812 features in the dynamic feature matrix extracted by the feature filter by using a Fisher algorithm, sorting the dynamic feature matrix according to the sequence of scores from top to bottom, and screening the first 80 features with high scores as the most representative features;
the data trainer is used for training the first 80 most representative features screened out according to the functional nuclear magnetic resonance image in the classification of the support vector machine to obtain a classifier;
the auxiliary diagnosis device is used for carrying out auxiliary diagnosis on the Alzheimer disease according to the trained classifier.
A method for using an Alzheimer disease auxiliary diagnosis device based on a dynamic brain function network comprises the following steps:
step 1: the preprocessing of the functional nuclear magnetic resonance image comprises the following steps:
1.1) carrying out time slice correction on the obtained 606 functional nuclear magnetic resonance images to be detected through the time slice corrector to obtain 606 time slice corrected functional nuclear magnetic resonance images (I)-1,I-2,I-3,…,I-606)；
1.2) performing head motion correction on the 606 time-corrected functional nuclear magnetic resonance images through the head motion corrector to obtain 606 head motion-corrected functional nuclear magnetic resonance images (H)-1,H-2,H-3,…,H-606)；
1.3) carrying out spatial standardization on the 606 functional nuclear magnetic resonance images after the head movement correction through the spatial standardizer to obtain 606 functional nuclear magnetic resonance images (F) after the spatial standardization-1,F-2,F-3,…,F-606)；
1.4) the 606 spatial normalized functional nuclear magnetic resonance images are subjected to smooth noise reduction through the smooth noise reducer to obtain 606 standard functional nuclear magnetic resonance images (S)-1,S-2,S-3,...,S-606)。
Step 2: the method for constructing the dynamic brain function network by utilizing the preprocessed image comprises the following steps of:
2.1) matching each standard functional nuclear magnetic resonance image in the 606 standard functional nuclear magnetic resonance images with a standard power-264 template of 264 brain areas through the template matcher, wherein each matched functional nuclear magnetic resonance image comprises 264 brain areas, and obtaining 606 matched functional nuclear magnetic resonance images (A)-1,A-2,A-3,...,A-606)；
2.2) passing 606 matched functional nuclear magnetic resonance images through the time sequence divider to each brain region in each matched functional nuclear magnetic resonance imageIs divided into 25 time series of transients, each representing a temporal information (T) of a brain region-1,T-2,...,T-25) And 606 segmented functional nuclear magnetic resonance images are represented as
Each divided image is sequentially passed through a mutual information calculator, the matching relation among all nodes of the divided image is calculated, the representation of edges in each dynamic brain function network is determined, the correlation relation among all nodes is calculated by adopting a mutual information method, and an incidence matrix is formed by the correlation relation;
2.3) constructing 15150 dynamic brain function networks of 606 segmented functional nuclear magnetic resonance images by the mutual information value obtained by calculation through the brain network constructor
Calculating mutual information of any two nodes of the fMRI image, obtaining a correlation matrix, setting a threshold value of the mutual information to be 0.18, comparing a mutual information value with the set threshold value of 0.18, setting an edge of the mutual information value to be 1 when the mutual information value is larger than the set threshold value of 0.18, namely, the two nodes are related, or setting the edge of the mutual information value to be 0, namely, the edge is not related, and thus, converting the correlation matrix into an adjacent matrix to construct a dynamic brain function network.
The embodiment is used for performing computer-aided diagnosis on functional nuclear magnetic resonance images to help doctors to perform diagnosis, wherein a flow chart of a method for extracting, filtering, screening and assisting diagnosis of dynamic information features is shown in fig. 3, and the specific steps comprise a step 3 and a step 4.
Step 3, calculating node measurement for the dynamic brain function networks, extracting the dynamic characteristics of each dynamic brain function network, and splicing the characteristic values of the filtered dynamic characteristics into a dynamic characteristic matrix, wherein the method comprises the following steps:
3.1) generating 15150 dynamic brain function networks of 606 segmented functional nuclear magnetic resonance images by the node measurement generator15150 node measures of dynamic brain function network, and calculates 28057800 characteristic values of dynamic characteristicsThe 28057800 dynamic features comprise 60600 global features and 27997200 local features, the number of the global features calculated by each dynamic brain function network is 4, and the number of the local features calculated by each dynamic brain function network is 7;
the 4 global features are respectively: clustering _ coefficients, charateristicpath length, Global efficiency, Transitivity;
the 7 local features are respectively: local effectiveness, degree, betwennesscentrality, Pagerank centrality, node strength, k-core centrality, and flow coefficient.
3.2) generating a time sequence by each group of characteristic values through a time sequence generator, 1122312 time sequences formed by 15150 dynamic brain function networks are represented asEach group of feature values comprises a global group of feature values and a local group of feature values, each global group of feature values comprises 25 feature values of dynamic brain function networks of each global feature in each segmented functional nuclear magnetic resonance image, and each local group of feature values comprises 25 feature values of dynamic brain function networks obtained by grouping each local feature of 46200 local features in each segmented functional nuclear magnetic resonance image according to 264 brain regions;
the method comprises the steps of constructing 25 brain function networks by a time sequence divider and a brain network builder for one functional nuclear magnetic resonance image, calculating global features and local features of each formed brain network, and generating a time sequence with 25 time points by a time sequence generator under the 25 brain function networks for each feature.
3.3) the time series generated by each group of features is subjected to feature extraction again through the wavelet-based time series entropy in the feature extractor to obtain new feature values of 1122312 features
Features are extracted again based on the time series entropy of the wavelet. The concept of entropy is used in thermodynamics, after all, the word is a side of a fire word and is used for measuring the unavailability degree of energy of a system, and the larger the entropy is, the larger the unavailability degree of the energy is; the less energy is not available. Its physical meaning is a measure of the degree of misordering or complexity in the hierarchy. The application of entropy is also expanding, and the application is from thermodynamics to biology, physics and time series analysis. Shannon entropy (entropy) is a mathematically rather abstract concept, which can be understood as the probability of occurrence of a certain specific information (discrete random event). The more ordered a system is, the lower the information entropy is; conversely, the more chaotic a system is, the higher the entropy of the information becomes. Entropy of information is also a measure of the degree of system ordering
3.4) filtering 27997200 local features in the node metric by a feature filter, firstly calculating the degree of each brain region in each matched functional nuclear magnetic resonance image by a degree method in the node metric, and calculating the average value of the degrees of 264 brain regions in each matched functional nuclear magnetic resonance image
Then calculating the standard deviation sigma of 264 brain areas in each matched functional nuclear magnetic resonance image, and then enabling the degree in the 264 brain areas to belong to an interval
The brain areas are filtered into key brain areas, and finally, 59388 local characteristic values generated by the local characteristics of 8484 brain areas obtained by filtering 606 matched functional nuclear magnetic resonance images are filteredThe new eigenvalues generated with 2424 global signatures are spliced into a dynamic signature matrix by the signature filter.
And 4, performing auxiliary diagnosis on the Alzheimer disease by using the generated dynamic feature matrix, wherein the auxiliary diagnosis method comprises the following steps:
4.1) scoring 61812 features in the dynamic feature matrix extracted by the feature filter through a Fisher algorithm in a feature filter, sorting the features according to the sequence of scores from top to bottom, and screening the first 80 features with high scores as the most representative features;
the dynamic brain function network constructed by all functional nuclear magnetic resonance images is used for combining the characteristics extracted from the time sequence generated by all node measurement into a matrix, and the characteristic screening is carried out by a Fisher algorithm, wherein the Fisher criterion basic principle is to find a most appropriate projection axis, so that the distance between the projections of two types of samples on the axis is as far as possible, and the projection of each type of sample is as compact as possible, thereby leading the classification effect to be optimal.
4.2) training the screened first 80 most representative features through a support vector machine in a data trainer to obtain a classifier;
4.3) using an auxiliary diagnosis device to perform auxiliary diagnosis of the Alzheimer disease through the obtained classifier.
Claims (10)
1. An Alzheimer disease auxiliary diagnosis device based on a dynamic brain function network is characterized by comprising an fMRI data preprocessing unit, a dynamic brain function network construction unit, a feature unit for generating training and an svm classification auxiliary diagnosis unit, wherein I acquired functional nuclear magnetic resonance images to be tested are subjected to preprocessing by the fMRI data preprocessing unit to obtain i preprocessed standard functional nuclear magnetic resonance images, the i preprocessed standard functional nuclear magnetic resonance images are subjected to time dimension division by the dynamic brain function network construction unit to construct i s dynamic brain function networks of the i functional nuclear magnetic resonance images, the constructed i s dynamic brain function networks are subjected to transverse extraction of dynamic features of each dynamic brain function network by the feature unit for generating training, and the dynamic features extracted from the i s dynamic brain function networks and filtered by a feature filter are extracted And finally, performing feature screening by using a Fisher algorithm through the svm classification auxiliary diagnosis unit, screening representative features, and training for auxiliary diagnosis.
2. The dynamic brain function network-based auxiliary diagnostic apparatus for alzheimer's disease as claimed in claim 1, wherein the fMRI data preprocessing unit comprises a time slice corrector, a motion corrector, a spatial normalizer and a smooth noise reducer, and the i time slice corrected functional nuclear magnetic resonance images are obtained by time slice correcting the i time slice-corrected functional nuclear magnetic resonance images by the time slice corrector, then the i time corrected functional nuclear magnetic resonance images are subjected to motion correction by the motion corrector to obtain i motion corrected functional nuclear magnetic resonance images, then the i motion corrected functional nuclear magnetic resonance images are subjected to spatial normalization by the spatial normalizer to obtain i spatially normalized functional nuclear magnetic resonance images, and finally the i spatially normalized functional nuclear magnetic resonance images are subjected to smooth noise reduction by the smooth noise reducer Carrying out noise to obtain i standard functional nuclear magnetic resonance images;
the time slice corrector is used for performing time slice correction on the input I functional nuclear magnetic images to be detected to obtain I corrected functional nuclear magnetic resonance images (I) of time slices-1,I-2,I-3,…,I-i) Wherein i represents the number of the selected functional nuclear magnetic resonance images to be detected;
the head movement corrector is used for correcting the functional nuclear magnetic image (I) after the I time slices-1,I-2,I-3,...,I-i) Performing head motion correction to obtain i functional nuclear magnetic resonance images (H) after head motion correction-1,H-2,H-3,...,H-i)；
The spatial normalizer is used for correcting i functional nuclear magnetic images (H) after head movement-1,H-2,H-3,...,H-i) Spatially normalizing to obtain i spatially normalized functional nuclear magnetic resonance images (F)-1,F-2,F-3,...,F-i)；
The smoothing noise reducer is used for normalizing i spatially functional nuclear magnetic resonance images (F)-1,F-2,F-3,...,F-i) Performing smooth noise reduction to obtain i standard functional nuclear magnetic resonance images (S)-1,S-2,S-3,...,S-i)。
3. The Alzheimer's disease auxiliary diagnosis device based on dynamic brain function network according to claim 1, the method is characterized in that the dynamic brain function network building unit comprises a template matcher, a time sequence divider and a brain network builder, and comprises the steps of firstly, matching each standard function nuclear magnetic resonance image in i standard function nuclear magnetic resonance images with a standard matching template with the specification of m through the template matcher to obtain an interested region, enabling each matched function nuclear magnetic resonance image to comprise m brain regions, then, the matched functional nuclear magnetic resonance images are divided into the time sequence of each brain region through the time sequence divider to obtain s sections of instantaneous time sequences, and finally the s sections of instantaneous time sequences obtained through division are constructed into i x s dynamic brain function networks of the i divided functional nuclear magnetic resonance images through the brain network constructor;
the template matcher is used for matching each standard functional nuclear magnetic resonance image with a standard matching template with the specification of m, each matched functional nuclear magnetic resonance image comprises m brain areas, and the i matched functional nuclear magnetic resonance images are expressed as (A)-1,A-2,A-3,...,A-i)；
The time sequence divider is used for dividing the time sequence of m brain regions in each matched functional nuclear magnetic resonance image into s sections of instant time sequences, and each section of instant time sequence represents instant information (T) of one brain region-1,T-2,...,T-s)，The i segmented functional nuclear magnetic resonance images are represented as
Wherein the value range of s is determined according to a preset segmentation interval;
the brain network builder is used for building s dynamic brain function networks from s sections of instantaneous time sequences in each segmented functional nuclear magnetic resonance image, and the i segmented functional nuclear magnetic resonance images obtain i x s dynamic brain function networks
4. The Alzheimer's disease auxiliary diagnosis device based on dynamic brain function network according to claim 1, the method is characterized in that the feature unit for training comprises a node measurement generator, a time sequence generator, a feature extractor and a feature filter, i x s dynamic brain function networks for constructing i segmented functional nuclear magnetic resonance images output by the dynamic brain function network unit calculate node measurement of the i x s dynamic brain function networks through the node measurement generator, then, each node metric in the i x s dynamic brain function networks is formed into a time sequence through a time sequence generator, secondly, extracting new characteristic values for the time sequence formed by measuring each node through the characteristic extractor, and finally splicing all the extracted new characteristic values into a dynamic characteristic matrix after filtering through the characteristic filter;
the node metric generator is used for generating node metrics of i-s dynamic brain function networks and calculating the characteristic value of z dynamic characteristics to be represented as
The z dynamic features include i s x global features and i s m y local features, i.e., z i x + i s m y, x represents the number of global features calculated by each dynamic brain function network, and y represents the local features calculated by each dynamic brain function networkThe number of (2);
the time sequence generator is used for generating a time sequence from each group of obtained characteristic values, and z/s time sequences formed by i-s dynamic brain function networks are represented as
Each group of feature values comprises global feature values of each group and local feature values of each group, each global feature value of each group comprises feature values of s dynamic brain function networks of each global feature in each segmented functional nuclear magnetic resonance image, and each local feature value of each group comprises feature values of s dynamic brain function networks obtained by grouping each local feature of s m y local features in each segmented functional nuclear magnetic resonance image according to m brain regions;
the feature extractor is used for extracting features again for the time series generated by each group of features based on the time series entropy of the wavelet to obtain new feature values of z/s features
The feature filter is used for filtering i s m y local features in the node metric, firstly, the degree of each brain area in each matched functional nuclear magnetic resonance image is calculated by a degree method in the node metric, and the average value of m brain area degrees in each matched functional nuclear magnetic resonance image is calculated
Then calculating the standard deviation sigma of m brain areas in each matched functional nuclear magnetic resonance image, and then enabling the degree in the m brain areas to belong to an interval
Filtering the brain areas into key brain areas, and finally filtering the i matched functional nuclear magnetic resonance images to obtain new characteristic values generated by the N x y local characteristics of the N brain areas and new characteristics generated by the i x global characteristicsThe values are spliced into a dynamic feature matrix by a feature filter,
wherein n isjThe number of key brain areas in the jth matched functional nuclear magnetic resonance image is shown, and the number of matched functional nuclear magnetic resonance images is shown as i.
5. The dynamic brain function network-based Alzheimer disease auxiliary diagnosis device according to claim 1, wherein the svm classification auxiliary diagnosis unit comprises a feature filter, a data trainer and an auxiliary diagnostor, and the dynamic feature matrix generated by the output of the feature unit for training is firstly passed through the feature filter, the dynamic feature matrix is subjected to dynamic feature screening by using Fisher algorithm, then the classification training is carried out by the data trainer, and finally the diagnosis of Alzheimer disease is realized by the auxiliary diagnosticator;
the feature filter is used for scoring N x y + i x features in the dynamic feature matrix by using a Fisher algorithm, sorting the N x y + i x features according to the sequence of scores from top to bottom, screening the top w features with high scores as the most representative features, and determining w according to actual conditions;
the data trainer is used for training the screened first w most representative features according to the functional nuclear magnetic resonance image to obtain a classifier in the classification of the support vector machine;
the auxiliary diagnosis device is used for carrying out auxiliary diagnosis on the Alzheimer disease according to the trained classifier.
6. The method for using the dynamic brain function network-based Alzheimer's disease auxiliary diagnosis device of any one of claims 1-5, which comprises the following steps:
step 1: preprocessing a functional nuclear magnetic resonance image;
step 2: constructing a dynamic brain function network by utilizing the preprocessed image;
and step 3: calculating node measurement of the dynamic brain function networks, extracting dynamic characteristics of each dynamic brain function network, and splicing characteristic values of the filtered dynamic characteristics into a dynamic characteristic matrix;
and 4, step 4: and performing auxiliary diagnosis on the Alzheimer disease by using the generated dynamic feature matrix.
7. The method for using the dynamic brain function network-based Alzheimer's disease auxiliary diagnosis device according to claim 6, wherein the step 1 of preprocessing the functional nuclear magnetic resonance image comprises the following steps:
1.1) the acquired I functional nuclear magnetic resonance images to be detected are subjected to time slice correction by the time slice corrector to obtain I time slice corrected functional nuclear magnetic resonance images (I)-1,I-2,I-3,…,I-i) Wherein i represents the number of the selected functional nuclear magnetic resonance images to be detected;
1.2) performing head motion correction on the i time-corrected functional nuclear magnetic resonance images through the head motion corrector to obtain i head motion-corrected functional nuclear magnetic resonance images (H)-1,H-2,H-3,...,H-i)；
1.3) the i functional nuclear magnetic resonance images after the head movement correction are subjected to spatial standardization through the spatial standardizer to obtain i functional nuclear magnetic resonance images (F) after the spatial standardization-1,F-2,F-3,...,F-i)；
1.4) the i spatially normalized functional nuclear magnetic resonance images are subjected to smooth noise reduction by the smooth noise reducer to obtain i standard functional nuclear magnetic resonance images (S)-1,S-2,S-3,...,S-i)。
8. The method for using the dynamic brain function network-based Alzheimer's disease auxiliary diagnosis device according to claim 6, wherein the step 2 of constructing the dynamic brain function network by using the preprocessed images comprises the following steps:
2.1) matching each standard functional nuclear magnetic resonance image in the i standard functional nuclear magnetic resonance images with a standard matching template with the specification of m through the template matcher to obtain an interested area, wherein each matched functional nuclear magnetic resonance image comprises m brain areas, and i matched functional nuclear magnetic resonance images (A)-1,A-2,A-3,...,A-i)；
2.2) the i matched functional nuclear magnetic resonance images are used for segmenting the time sequence of each brain area in each matched functional nuclear magnetic resonance image through the time sequence segmenter to obtain s sections of instant time sequences, and each section of instant time sequence represents instant information (T) of one brain area-1,T-2,...,T-s) I segmented functional NMR images are expressed asWherein the value range of s is determined according to a preset segmentation interval;
2.3) constructing i x s dynamic brain function networks of i segmented functional nuclear magnetic resonance images by the segmented s-segment instantaneous time sequence through the brain network constructor
9. The method for using the dynamic brain function network-based alzheimer's disease auxiliary diagnosis device according to claim 6, wherein the step 3 calculates node metrics for the dynamic brain function networks, extracts dynamic features of each dynamic brain function network, and then concatenates the feature values of the filtered dynamic features into a dynamic feature matrix, comprising the following steps:
3.1) generating node measurement of i x s dynamic brain function networks of the i segmented functional nuclear magnetic resonance images through the node measurement generator, and calculating to obtain the characteristic value of z dynamic characteristics to be represented as
The z dynamic features comprise i s x global features and i s m y local features, namely z i s x i m y, x represents the number of global features calculated by each dynamic brain function network, and y represents the number of local features calculated by each dynamic brain function network;
3.2) generating a time sequence by the time sequence generator for each group of characteristic values, and then expressing z/s time sequences formed by i × s dynamic brain function networks as
Each group of feature values comprises global feature values of each group and local feature values of each group, each global feature value of each group comprises feature values of s dynamic brain function networks of each global feature in each segmented functional nuclear magnetic resonance image, and each local feature value of each group comprises feature values of s dynamic brain function networks obtained by grouping each local feature of s m y local features in each segmented functional nuclear magnetic resonance image according to m brain regions;
3.3) the time series generated by each group of features is subjected to feature extraction again through the wavelet-based time series entropy in the feature extractor to obtain new feature values of z/s features
3.4) filtering i s m y local features in the node measure by a feature filter, firstly calculating the degree of each brain area in each matched functional nuclear magnetic resonance image by a degree method in the node measure, and calculating the average value of the degrees of the m brain areas in each matched functional nuclear magnetic resonance image
Then calculating the standard deviation sigma of m brain areas in each matched functional nuclear magnetic resonance image, and then enabling the degree in the m brain areas to belong to an interval
Filtering the brain areas to obtain key brain areas, finally splicing new eigenvalues generated by N x local characteristics of the N brain areas and i x global characteristics obtained by filtering the matched functional nuclear magnetic resonance images into a dynamic characteristic matrix through a characteristic filter,wherein n isjThe number of key brain areas in the jth matched functional nuclear magnetic resonance image is shown, and the number of matched functional nuclear magnetic resonance images is shown as i.
10. The method for using the dynamic brain function network-based Alzheimer's disease auxiliary diagnosis device according to claim 6, wherein the step 4 of performing auxiliary diagnosis on Alzheimer's disease by using the generated dynamic feature matrix comprises the following steps:
4.1) scoring N x y + i x features in the dynamic feature matrix extracted by the feature filter through a Fisher algorithm in a feature filter, sorting the features according to the sequence of scores from top to bottom, screening the top w features with high scores as the most representative features, and determining w according to the actual condition;
4.2) training the first w most representative features screened out by a support vector machine in a data trainer to obtain a classifier;
4.3) using an auxiliary diagnosis device to perform auxiliary diagnosis of the Alzheimer disease through the obtained classifier.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910916563.8A CN110718301B (en) | 2019-09-26 | 2019-09-26 | Alzheimer's disease auxiliary diagnosis device and method based on dynamic brain function network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910916563.8A CN110718301B (en) | 2019-09-26 | 2019-09-26 | Alzheimer's disease auxiliary diagnosis device and method based on dynamic brain function network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110718301A true CN110718301A (en) | 2020-01-21 |
| CN110718301B CN110718301B (en) | 2024-04-19 |
Family
ID=69210922
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910916563.8A Active CN110718301B (en) | 2019-09-26 | 2019-09-26 | Alzheimer's disease auxiliary diagnosis device and method based on dynamic brain function network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110718301B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112991335A (en) * | 2021-04-23 | 2021-06-18 | 合肥综合性国家科学中心人工智能研究院(安徽省人工智能实验室) | Imagination capacity evaluation method and system and readable storage medium |
| CN113177633A (en) * | 2021-04-20 | 2021-07-27 | 浙江大学 | Deep decoupling time sequence prediction method |
| CN116541751A (en) * | 2023-07-03 | 2023-08-04 | 中国医学科学院生物医学工程研究所 | Electroencephalogram signal classification method based on brain function connection network characteristics |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110301431A1 (en) * | 2010-06-05 | 2011-12-08 | The Board Of Trustees Of The Leland Stanford Junior University | Methods of classifying cognitive states and traits and applications thereof |
| US20130231552A1 (en) * | 2012-03-05 | 2013-09-05 | Siemens Corporation | Method and System for Diagnosis of Attention Deficit Hyperactivity Disorder from Magnetic Resonance Images |
| CN104715261A (en) * | 2015-03-23 | 2015-06-17 | 南京工业大学 | fMRI dynamic brain function sub-network construction and parallel SVM weighted recognition method |
| CN107256408A (en) * | 2017-05-12 | 2017-10-17 | 常州大学 | A kind of critical path searching method of cerebral function network |
| CN109859839A (en) * | 2019-01-23 | 2019-06-07 | 桂林电子科技大学 | A kind of Alzheimer's disease classification method based on layering integrated study |
-
2019
- 2019-09-26 CN CN201910916563.8A patent/CN110718301B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110301431A1 (en) * | 2010-06-05 | 2011-12-08 | The Board Of Trustees Of The Leland Stanford Junior University | Methods of classifying cognitive states and traits and applications thereof |
| US20130231552A1 (en) * | 2012-03-05 | 2013-09-05 | Siemens Corporation | Method and System for Diagnosis of Attention Deficit Hyperactivity Disorder from Magnetic Resonance Images |
| CN104715261A (en) * | 2015-03-23 | 2015-06-17 | 南京工业大学 | fMRI dynamic brain function sub-network construction and parallel SVM weighted recognition method |
| CN107256408A (en) * | 2017-05-12 | 2017-10-17 | 常州大学 | A kind of critical path searching method of cerebral function network |
| CN109859839A (en) * | 2019-01-23 | 2019-06-07 | 桂林电子科技大学 | A kind of Alzheimer's disease classification method based on layering integrated study |
Non-Patent Citations (3)
| Title |
|---|
| 皇甫浩然;杨剑;杨阳;: "基于fMRI动态功能连接的抑郁症患者分类研究", 计算机应用研究, no. 03 * |
| 董国昭;曾晓天;杨柳;张逸鹤;张勇;唐晓英;: "基于动态图论特征的阿尔茨海默病早期预测", 北京生物医学工程, no. 06 * |
| 马士林;梅雪;李微微;周宇;: "fMRI动态功能网络构建及其在脑部疾病识别中的应用", 计算机科学, no. 10 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113177633A (en) * | 2021-04-20 | 2021-07-27 | 浙江大学 | Deep decoupling time sequence prediction method |
| CN113177633B (en) * | 2021-04-20 | 2023-04-25 | 浙江大学 | Depth decoupling time sequence prediction method |
| CN112991335A (en) * | 2021-04-23 | 2021-06-18 | 合肥综合性国家科学中心人工智能研究院(安徽省人工智能实验室) | Imagination capacity evaluation method and system and readable storage medium |
| CN112991335B (en) * | 2021-04-23 | 2024-04-26 | 合肥中聚源智能科技有限公司 | Imagination capability assessment method, system and readable storage medium |
| CN116541751A (en) * | 2023-07-03 | 2023-08-04 | 中国医学科学院生物医学工程研究所 | Electroencephalogram signal classification method based on brain function connection network characteristics |
| CN116541751B (en) * | 2023-07-03 | 2023-09-12 | 中国医学科学院生物医学工程研究所 | Electroencephalogram signal classification method based on brain function connection network characteristics |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110718301B (en) | 2024-04-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Li et al. | Epileptic seizure detection in EEG signals using sparse multiscale radial basis function networks and the Fisher vector approach | |
| CN110188836B (en) | Brain function network classification method based on variational self-encoder | |
| CN110718301A (en) | Alzheimer disease auxiliary diagnosis device and method based on dynamic brain function network | |
| CN110705456A (en) | Micro motor abnormity detection method based on transfer learning | |
| JP2004174254A5 (en) | ||
| CN111009324B (en) | Auxiliary diagnosis system and method for mild cognitive impairment through multi-feature analysis of brain network | |
| CN101794372A (en) | Method for representing and recognizing gait characteristics based on frequency domain analysis | |
| Naresh et al. | Early detection of lung cancer using neural network techniques | |
| KR20160049897A (en) | Computer aided diagnosis apparatus and method using a sequence of medical images | |
| EP2987114B1 (en) | Method and system for determining a phenotype of a neoplasm in a human or animal body | |
| CN111783887B (en) | Classified lie detection identification method based on fMRI (magnetic resonance imaging) small-world brain network computer | |
| CN112465905A (en) | Characteristic brain region positioning method of magnetic resonance imaging data based on deep learning | |
| CN104838422A (en) | Image processing device and method | |
| Tun et al. | Feature extraction and classification of lung cancer nodule using image processing techniques | |
| CN115272295A (en) | Dynamic brain function network analysis method and system based on time domain-space domain combined state | |
| CN113069117A (en) | Electroencephalogram emotion recognition method and system based on time convolution neural network | |
| Thomas et al. | Artificial neural network for diagnosing autism spectrum disorder | |
| CN108805181B (en) | Image classification device and method based on multi-classification model | |
| CN112336369B (en) | Coronary heart disease risk index evaluation system of multichannel heart sound signals | |
| Mousavian et al. | Depression detection using atlas from fMRI images | |
| CN116543154A (en) | Medical image segmentation method based on multi-level semantic features | |
| Dinu et al. | A novel modelling technique for early recognition and classification of Alzheimer’s disease | |
| CN110647870A (en) | Method for calculating approximate entropy of resting state fMRI data based on sliding window | |
| AltundoĞan et al. | Cracked Wall Image Classification Based on Deep Neural Network Using Visibility Graph Features | |
| EP4198997A1 (en) | A computer implemented method, a method and a system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |